The FCA just called out rubber-stamp human-in-the-loop controls
The line in the FCA's new paper that should worry boards most is the one about human-in-the-loop. In some firms, reviewers are accepting AI recommendations without scrutiny, which means the control everyone points to in the governance committee isn't really a control. I've sat in that kind of review. The slide says a human reviews every decision, and nobody asks how often that human disagrees with the model. If the override rate is near zero, you don't have oversight, you have a signature. Under SM&CR, a named senior manager is personally accountable for that. So what's your override rate, and could you prove the reviewer was actually looking?
Source: FCA finds AI governance lagging behind deployment across financial firms - Insurance Business (Insurance Business)